How a Dutch Company Became a Tiny Bottleneck in the Enormous Semiconductor Supply Problem

semiconductor shortage graphic

The shortage of semiconductors has had a profound effect on the global economy. Semiconductors, in particular microprocessors (semiconductors containing a CPU, essentially a computer), are the brains inside so many things that are a part of daily life: washing machines, toys, watches, cameras, ATMs, planes, televisions, LED light bulbs and nearly every form of human communication. We don’t really drive cars anymore, we basically drive computers. You use semiconductor chips constantly, and they enable your life to be much easier. In 2019, nearly 300 billion chips were sold, which is 40 chips for every person on the planet.

This is owed in large part to the idea behind Moore’s Law, which essentially said that computing power would double in speed and costs cut in half every two years or so. What Gordon Moore, the co-founder of Intel, actually said was that the number of transistors in a dense integrated circuit would double every two years. This has held true since the essay behind Moore’s Law was published in 1965. By 1971, Intel would produce the first microprocessor, which had 2,300 transistors. Today’s chips, no larger than the size of a person’s fingernail, each contain billions of transistors.

The development of semiconductors this small and this powerful is a technological marvel, and its value to the industry, to people, really can’t be overstated. It is, without question, humankind’s greatest manufacturing achievement. Only our relentlessly increasing demand for more and more semiconductors across nearly every industry could have driven such advancement in such a short amount of time. Getting to this point was incredibly expensive and difficult, taking the world’s most brilliant minds many years of research and development.

Everything about semiconductor production at scale is complex and expensive. The upfront costs of building new foundries are enormous. A lesser-known companion to Moore’s Law, Rock’s Law, states that the cost to build a semiconductor chip fabrication plant doubles every four years. Every part of the production process must take place in a clean room, with workers wearing full suits, to protect the delicate wafers from dust, dirt or any of the many tiny particles that can ruin them. There are many pieces of critical equipment used in the manufacturing process, all expensive and incredibly complicated in and of themselves, of course. How the chips are made, sorted and tested are all critical pieces of the process. Market demand requires smaller, faster, cheaper chips. That’s only possible if the yield of new chips coming from the foundries increases. Advancement in the manufacturing processes is as important, if not actually more important, as the advancement of chip design.

Arguably the most important piece of the puzzle is the manufacturing technique that makes it possible to even create chips with billions of transistors: photolithography. Photolithography is the process in which light is used to etch intricate designs onto very thin silicon wafers. The light is passed through optical lenses in order to shrink the pattern. The more intricate and detailed the designs, the more transistors that can fit on each chip.

The key to getting more transistors on chips is the wavelength of the light used in the photolithography, with shorter wavelengths equaling more transistors. Until a few years ago, the most advanced option that had commercial viability was deep-ultraviolet (DUV) light, which has a wavelength of 193 nanometers. Transistors can only get so small using this kind of light, and there is significant demand for them to continue getting smaller.

Companies have poured billions of dollars and decades of work into research and development for next-generation photolithography, in what are certainly the most complex manufacturing problems that have ever existed. There are actually only three companies that make and supply semiconductor lithography equipment at all: Canon, Nikon and ASML, a Dutch company spun off of Philips in 1984. Canon and Nikon both abandoned working on new photolithography techniques, instead working on ways to keep advancing DUV. In the end, only one next-gen technique became commercially viable: extreme ultraviolet lithography (EUVL), and only one company broke through the technology barriers to bring it to market: ASML.

EUVL works in a similar manner to deep-ultraviolet, except that mirrors are used instead of optical lenses, to create a reflective mask. It also has a much shorter wavelength of 13.5 nm, significantly shorter than previous techniques. Getting EUVL to market has been extremely challenging.

ASML pushed forward with EUVL, and after 20 years and billions in investment, finally solved the biggest impediment to its manufacturing success: Previously, there was no source that produced enough EUV light to complete the process. At least, not on earth — EUV light exists in outer space, but ASML needed to create and control it not just here on earth, but within a closed manufacturing system with tightly controlled parameters. The answer to the light source problem was laser-produced-plasma (LPP) at 250W, which was enough power to achieve the necessary “wafer throughput.”

ASML finally sold its first commercial EUVL machine in 2018 to Samsung. It currently has 100% market share for EUVL machinery. Although many chips are made using other types of photolithography equipment, EUVL is the most cutting edge technology available, and the only option for the most advanced semiconductor chips.

In addition to being the exclusive provider of EUVL machines, ASML also made a significant breakthrough in DUV technology that allowed two parts of the process in the etching phase to be performed simultaneously, which doubled manufacturing speed. Although Nikon still sells DUV photolithography equipment, ASML is the market leader in that category as well.

A single EUVL machine from ASML costs upwards of $150 million, and is so large that it takes 40 40-foot shipping containers to ship. There are now more than 100 EUVL machines operating in semiconductor foundries around the world, but demand keeps increasing. There is a long waiting list.

With such huge capital costs for opening new semiconductor foundries, and the time and effort it takes to get such facilities going, the semiconductor industry relies heavily upon accurate forecasting and demand planning to understand what the market will demand in future years. Unprecedented events like the COVID-19 pandemic have thrown all forecasts out the window. At first, the COVID-19 pandemic caused demand to drop, as consumers stopped buying stuff and companies who sell to them immediately cut back on production to save money. But then, with people retreating to their homes and needing more semiconductor-driven devices to work, socialize and entertain themselves from the safety of their homes, demand didn’t just rebound, it soared in 2021. Work productivity decreased, at least for manufacturing and the supply chain industry, which has few jobs that can be done remotely. Lead time for semiconductor orders increased, which has been one of the biggest drivers in the shortage of other products, everything from cars to video game consoles, and indeed, commercial and consumer audiovisual equipment.

It’s disconcerting to understand that so few companies, and in particular one, sit in a position to so drastically affect the global economy. It took ASML almost 20 years to bring EUVL to market, and so much money (there’s no official total figure, but it is billions per year). They could have easily justified abandoning the effort, like Nikon and Canon did. Many experts actually called them crazy for not doing so. Thankfully, they persevered. Now, ASML is worth $265 billion and one of Europe’s largest companies. It might be a good time to think about buying some ASML stock.